An escalation path specifies the conditions triggering human review: low confidence scores, high-stakes determinations, conflicting sources, or categories designated for mandatory review. It identifies the reviewer and the response timeframe.
It is a component of AI governance and workflow design.
Alternative Names:
Escalation Procedure, Human Escalation
Why it Matters?
Escalation design is what makes supervision obligations operational rather than aspirational, since a policy stating that attorneys review AI output means little without defining which output, by whom, and when. Effective paths route by consequence rather than volume: a privilege determination escalates regardless of confidence, while a document type classification may not. Documenting the escalation criteria is also what demonstrates supervision if it is later questioned.
Frequently Confused with
Related terms
Frequently asked questions
What triggers escalation?
How should paths be designed?
